Analytics preference

We only load Google Analytics after you opt in. You can change this choice at any time from Cookie settings.

Expression Evaluation Dialog

The Expression Evaluation Dialog is the expanded editor for expression-capable fields.

The Expression Evaluation Dialog is the expanded editor for expression-capable fields. It opens from the expand button next to an expression input and gives you a larger workspace, autocomplete, live backend preview, and output browsing in one place.

Opening the Dialog

  • Click the expand button next to an expression field in the Properties panel.
  • Multi-line fields show the button inside the input.
  • Single-line fields show the button to the right of the input.

Dialog Layout

  • The dialog opens as a centered modal at roughly 80vw × 80vh.
  • The header shows Evaluate — {field name} when the field label is known.
  • The editor keeps autocomplete, field-to-field navigation, and the usual Apply / Cancel actions.
  • The preview refreshes automatically about 300 ms after you stop typing.
  • Press Ctrl / Cmd + Enter to apply the current text back to the field and close the dialog (same as Apply).

Output Section

The Output area refreshes automatically after you pause typing in the editor.

Evaluation

Expression typeResult
Single $expr (entire field)Native value is preserved
Single bare dot path (entire field, no $, no embedded newline)Same as $ + that path (backend normalizes for evaluation)
Text with embedded $refsFinal result is always a string
Literal text (no $, not a single dot path)Returned as-is

Examples

InputResult
$input.items[1, 2, 3]
input.items (same as $input.items when the whole value is this path)[1, 2, 3]
Hello $input.nameHello John
$input.count + 16
just textjust text

Evaluation context is built in this order:

  • Pinned data on upstream nodes
  • Last run outputs from the canvas
  • null when no upstream data is available

If the current node has upstream dependencies but none of them have pinned data or a recent run result, the dialog first performs a test run to build context and then evaluates the current text.

Output Path Picker

When the backend result is an object or array, the Output area can show an interactive path picker instead of raw JSON.

  • Double-click a row to append that path to the current $node.path reference.
  • Single-click expandable rows to open or close branches.
  • Use the built-in query box to search keys, values, and path strings.

Autocomplete

  • Type $ to open suggestions for upstream nodes, built-ins, and methods.
  • Press Tab or Enter to insert the selected suggestion.
  • Use ↑ / ↓ to move through the suggestion list.

Example Answers

When you continue a $… reference with a dot, each suggestion row shows the answer that suggestion would produce, in bold on the right of the row.

  • Typing $set.listItems.first().toString().u lists upper, urlEncode, urlDecode, and unescape, each with its evaluated result — so you can pick by outcome instead of by name.
  • Typing $node.text.toJson() parses a JSON string into an object or array. After .toJson(), autocomplete can offer the parsed keys (for example repoUrl) when last-run or pinned output is available.
  • Results are evaluated on the backend against the same pinned data and last-run outputs the dialog preview uses, so they match what the field will produce.
  • Only suggestions that stand on their own are evaluated: zero-argument methods (upper(), first(), keys(), toJson()) and plain properties (length, year). Suggestions that need arguments first — map, filter, sort, replace — are left blank.
  • A row stays blank when the candidate produces no answer: an error, a null, or a method that does not apply to that value.
  • Answers are shortened to one line. Hover a shortened answer to see more of it.

Build with AI

The Build with AI button appears in the top-right corner of the evaluate dialog when a workflow is loaded. It opens a nested modal where you can describe the expression you want in plain text and have an LLM generate it for you.

How it works

  1. Click Build with AI in the evaluate dialog toolbar.
  2. Select an LLM credential and model from the dropdowns.
  3. Type a plain-text description of what you want — for example, "Get the customer name from the API response".
  4. Click Generate. The modal calls the backend, which uses the Expression DSL system prompt plus your last workflow run outputs as context, and returns a single expression string.
  5. The generated expression is immediately evaluated against the same canvas data so you can see the result before applying.
  6. Click Apply to replace the editor content with the generated expression, or Regenerate to try again.

Context used

The generator receives the node outputs from the last test run on the canvas. If the workflow has not been run yet, the LLM generates a best-effort expression based on node labels and DSL conventions.

Requirements

  • An LLM credential (OpenAI, Google, or Custom) must be configured in the Credentials tab.
  • The workflow must be open in the editor.

Keyboard Shortcuts

ShortcutAction
Typing pause (~300 ms)Refresh backend preview
Ctrl / Cmd + EnterApply and close the dialog
EscapeClose autocomplete first, otherwise close the dialog

Nested $ in method arguments

For dialog / backend evaluation, you can pass another reference with a nested $ inside calls such as .get($other.field) when you need a dynamic key. You can also use a bare path inside .get(...) (for example other.field); the backend resolves it the same way. Saved workflow fields should still follow the Expression DSL rules; the evaluate API also accepts a full-line bare dot path for convenience (see table above).

Set and Execute mapping editors can show Prev / Next buttons so you can move between fields without closing the dialog.